Genotype-Environment Interaction Concepts for Field Experimentation

Abstract

Experimental limitations restrict the extent to which an experimenter can sample time and area in quantitative genetic field studies. Yet the sampling of environments is extremely important to assess the value of a genotype. A reference set of experiments based on sets of random environments available in time and space was characterized and used as a framework for considering genotype-environment interactions. The nature of the biases introduced when data are analyzed as though they had been collected from a random sample of environments, when in fact they had been collected in a non-random fashion, was examined. The reference for discussion was the random environment concept for evaluating genotypes. The interpretation entailed an evaluation of a function involving 2 intraclass correlations. Intraclass correlation estimates for a range of crops were obtained from experiments conducted primarily in North Carolina. From the magnitude of the estimates obtained for the intraclass correlations and from an evaluation of biases, data collected in associated environments could be treated as arising from random environments without unduly affecting the estimates of genetic and genotype-environmental interaction variances. Unbalanced sampling of environments, such as 1 different locations each year, was evaluated as a method of reducing biases when such data were to be analyzed as though environments had been randomly samples.